Contending with contradictory data in a risk assessment context: the case of methylmercury.

Two large sample, prospective longitudinal studies one in the Seychelles Islands in the Indian Ocean, the other in the Faroes Islands in the North Sea were conducted during the 1990s to examine the effects of prenatal methylmercury exposure on intellectual function in childhood. The Faroes study found evidence linking this exposure to adverse outcome, but the Seychelles study did not. A peer review workshop held in Raleigh, NC, in 1998 concluded that the inconsistencies between the Faroes and Seychelles findings could be explained by differences in study design and sources of exposure. The US Environmental Protection Agency contracted with the National Academy of Sciences (NAS) to convene an expert panel to provide guidance for a new risk assessment for methylmercury. The NAS panel reviewed the Faroes and Seychelles studies in light of data from a smaller New Zealand study and other data not available to the Raleigh reviewers. These additional data provided evidence of adverse effects in studies whose design and source of exposure were similar to that in the Seychelles, leading the NAS panel to conclude that the weight of the evidence supported the Faroes findings. A power analysis, conducted by computing standardized regression coefficients for the three studies, indicated that many of the Faroes findings were so subtle that the power to detect them in the Seychelles study, despite its large sample size, was only about 50%. Because prospective epidemiological studies are often hampered by limited control over confounding and other factors, including unmeasured between cohort differences in genetic vulnerability and nutritional adequacy, inferences about toxicity often depend heavily on a qualitative assessment of the weight of the evidence from multiple studies.